n8n nested AI agent tool: one idempotency_key per Sume generation

Give each intended generation one idempotency_key and make every level of agents reuse it for that generation. Sume's MCP requires the key on every write and paid tool, and it is a stable key for transport and dedup. Two agent levels that each invent their own key for the same intent look like two requests.

The n8n fact is from its release notes; the Sume facts are from MCP tools and gates and Jobs and results, read 2026-09-30.

What did n8n change?

The release notes for n8n 2.42.0 list under AI Agent Node: "Let an AI Agent Tool use its own tools under a pre-v3 parent agent" (issue 39160). In practice a sub-agent can now hold tools, which may include one of the hosted Sume tools, beneath a parent agent.

Why does nesting matter for a paid tool?

More agent levels means more chances for a retry or a second planning pass to call the same create. The docs table is plain: idempotency_key is "Required on write and paid tools", described as a stable key for transport and dedup, not human approval.

How do I pass the key down?

Put the key in the sub-agent's input (for example, derived from your workflow's run and item), and tell the sub-agent in its prompt to use exactly that value. The Sume docs do not describe n8n, so that wiring is yours to build. Ask the tool for its schema first: tools_schema with the tool name explains idempotency_key and dry_run before a paid submit.

What if a level times out?

Re-read with jobs_wait and the same ids. The docs say a wait slice is bounded and to retry the wait, never the paid create. More on key behavior in idempotency keys for AI video APIs.